coderev-cli 1.0.26 → 1.3.0

package/src/rag-indexer.js

@@ -0,0 +1,700 @@
+/**
+ * RAG (Retrieval-Augmented Generation) Codebase Indexer
+ *
+ * Phase 1: Lightweight local codebase indexing with text-based symbol extraction
+ * and cosine-similarity retrieval using simple TF-IDF vectors with optional LLM embeddings.
+ *
+ * Design:
+ * - No native deps (tree-sitter, sqlite-vec) — pure JS for Phase 1
+ * - Index stored as JSON in `.coderev/index/`
+ * - Two modes:
+ *   1. Fast: TF-IDF on extracted symbols/functions (no LLM call, instant)
+ *   2. Embedded: Uses LLM embeddings API for semantic search (needs API key, more accurate)
+ * - Indexed content: function signatures, class definitions, import statements, type defs
+ * - Diff context retrieval: given a changed file, find related symbols from the same
+ *   file + cross-file references (imports/exports)
+ */
+
+const fs = require('fs');
+const path = require('path');
+const crypto = require('crypto');
+
+const INDEX_DIR = '.coderev/index';
+const INDEX_FILE = 'codebase-index.json';
+const META_FILE = 'index-meta.json';
+
+// File extensions to index
+const INDEXABLE_EXTS = new Set([
+  '.js', '.jsx', '.ts', '.tsx', '.mjs', '.cjs',
+  '.py', '.rs', '.go', '.java', '.rb', '.php',
+  '.swift', '.kt', '.kts', '.c', '.cpp', '.h', '.hpp',
+  '.cs', '.sql', '.yaml', '.yml', '.toml',
+  '.vue', '.svelte', '.astro',
+]);
+
+// Directories to skip
+const SKIP_DIRS = new Set([
+  'node_modules', '.git', '__pycache__', '.venv', 'venv',
+  'dist', 'build', '.next', '.nuxt', '.output',
+  'target', 'bin', 'obj', '.gradle', '.idea',
+  'vendor', 'coverage', '.coderev',
+]);
+
+// Regex patterns for symbol extraction (language-agnostic)
+const SYMBOL_PATTERNS = [
+  // JavaScript/TypeScript: function declarations & arrow functions
+  {
+    lang: 'js/ts',
+    re: /(?:export\s+)?(?:async\s+)?function\s+(\w+)\s*\(([^)]*)\)/g,
+    type: 'function',
+  },
+  // JavaScript/TypeScript: arrow functions
+  {
+    lang: 'js/ts',
+    re: /(?:const|let|var)\s+(\w+)\s*=\s*(?:async\s*)?(?:\([^)]*\)|[\w$]+)\s*=>/g,
+    type: 'function',
+  },
+  // JavaScript/TypeScript: class declarations
+  {
+    lang: 'js/ts',
+    re: /(?:export\s+)?(?:abstract\s+)?class\s+(\w+)(?:\s+extends\s+(\w+))?/g,
+    type: 'class',
+  },
+  // JavaScript/TypeScript: method definitions in classes/objects
+  {
+    lang: 'js/ts',
+    re: /(?:(?:async\s+)?(\w+)\s*\(([^)]*)\)\s*\{)/g,
+    type: 'method',
+  },
+  // JavaScript/TypeScript: imports
+  {
+    lang: 'js/ts',
+    re: /(?:import\s+(?:(?:\{[^}]*\}|\*\s+as\s+\w+|\w+)(?:\s*,\s*(?:\{[^}]*\}|\*\s+as\s+\w+|\w+))*\s+from\s+)?['"]([^'"]+)['"])|(?:require\s*\(\s*['"]([^'"]+)['"]\s*\))/g,
+    type: 'import',
+  },
+  // JavaScript/TypeScript: exports
+  {
+    lang: 'js/ts',
+    re: /export\s+(?:default\s+)?(?:(?:function|class|const|let|var)\s+)?(\w+)/g,
+    type: 'export',
+  },
+  // Python: function definitions
+  {
+    lang: 'python',
+    re: /(?:async\s+)?def\s+(\w+)\s*\(([^)]*)\)/g,
+    type: 'function',
+  },
+  // Python: class definitions
+  {
+    lang: 'python',
+    re: /class\s+(\w+)(?:\s*\(([^)]*)\))?:/g,
+    type: 'class',
+  },
+  // Python: imports
+  {
+    lang: 'python',
+    re: /(?:from\s+(\S+)\s+import\s+(\S+))|(?:import\s+(\S+))/g,
+    type: 'import',
+  },
+  // Go: function declarations
+  {
+    lang: 'go',
+    re: /func\s+(?:\([^)]+\)\s+)?(\w+)\s*\(([^)]*)\)/g,
+    type: 'function',
+  },
+  // Go: type/struct definitions
+  {
+    lang: 'go',
+    re: /type\s+(\w+)\s+(?:struct|interface)\s*\{/g,
+    type: 'type',
+  },
+  // Go: imports
+  {
+    lang: 'go',
+    re: /"([^"]+)"/g,
+    type: 'import',
+  },
+  // Rust: function definitions
+  {
+    lang: 'rust',
+    re: /(?:pub\s+)?(?:async\s+)?fn\s+(\w+)\s*(?:<[^>]*>)?\s*\(([^)]*)\)/g,
+    type: 'function',
+  },
+  // Rust: struct/enum/trait
+  {
+    lang: 'rust',
+    re: /(?:pub\s+)?(?:struct|enum|trait)\s+(\w+)/g,
+    type: 'type',
+  },
+  // Java/Kotlin: method declarations
+  {
+    lang: 'java/kotlin',
+    re: /(?:(?:public|private|protected)\s+)?(?:static\s+)?\w+\s+(\w+)\s*\(([^)]*)\)/g,
+    type: 'method',
+  },
+  // Java/Kotlin: class
+  {
+    lang: 'java/kotlin',
+    re: /(?:public\s+)?class\s+(\w+)/g,
+    type: 'class',
+  },
+];
+
+// Language detection by extension
+const EXT_LANG_MAP = {
+  '.js': 'js/ts', '.jsx': 'js/ts', '.ts': 'js/ts', '.tsx': 'js/ts',
+  '.mjs': 'js/ts', '.cjs': 'js/ts',
+  '.py': 'python', '.pyw': 'python',
+  '.rs': 'rust',
+  '.go': 'go',
+  '.java': 'java/kotlin', '.kt': 'java/kotlin', '.kts': 'java/kotlin',
+  '.rb': 'ruby',
+  '.php': 'php',
+  '.swift': 'swift',
+  '.c': 'c', '.cpp': 'c/cpp', '.h': 'c', '.hpp': 'c/cpp',
+  '.cs': 'csharp',
+  '.sql': 'sql',
+  '.vue': 'js/ts', '.svelte': 'js/ts', '.astro': 'js/ts',
+};
+
+/**
+ * Determine the language group for a file extension.
+ */
+function langForExt(ext) {
+  return EXT_LANG_MAP[ext] || 'generic';
+}
+
+/**
+ * Extract symbols from source code text.
+ * Returns an array of { name, type, signature, lang, line } objects.
+ */
+function extractSymbols(source, filename) {
+  const ext = path.extname(filename).toLowerCase();
+  const lang = langForExt(ext);
+  const symbols = [];
+  const lines = source.split('\n');
+
+  for (const pattern of SYMBOL_PATTERNS) {
+    if (!pattern.lang.includes(lang) && pattern.lang !== 'generic') continue;
+
+    // Reset lastIndex for new source
+    const re = new RegExp(pattern.re.source, pattern.re.flags);
+
+    let match;
+    while ((match = re.exec(source)) !== null) {
+      let name, signature;
+
+      if (pattern.type === 'import') {
+        // Import patterns: capture the module path
+        name = match[1] || match[2] || match[3] || match[4] || '';
+        // For Go multi-import blocks, filter noise
+        if (lang === 'go' && name.startsWith('"') && name.endsWith('"')) {
+          name = name.slice(1, -1);
+        }
+      } else {
+        // Function/class/method: first capture group is name
+        name = match[1];
+        signature = match[2] || '';
+      }
+
+      if (!name || name.length < 1) continue;
+
+      // Skip noise words (language-specific)
+      const noise = ['if', 'for', 'while', 'switch', 'catch', 'return', 'throw',
+        'typeof', 'instanceof', 'delete', 'void', 'else', 'case', 'default',
+        'break', 'continue', 'try', 'finally', 'debugger', 'with'];
+      if (noise.includes(name) && pattern.lang !== 'rust') continue;
+
+      // Calculate line number
+      const pos = match.index;
+      const line = source.substring(0, pos).split('\n').length;
+
+      symbols.push({
+        name,
+        type: pattern.type,
+        signature: signature || '',
+        lang,
+        file: filename,
+        line,
+        // Context snippet for retrieval
+        snippet: lines.slice(Math.max(0, line - 2), Math.min(lines.length, line + 3)).join('\n'),
+      });
+    }
+  }
+
+  return symbols;
+}
+
+/**
+ * Simple TF-IDF style tokenizer for text.
+ */
+function tokenize(text) {
+  return text
+    .toLowerCase()
+    .replace(/[^a-z0-9_$]/g, ' ')
+    .split(/\s+/)
+    .filter(t => t.length > 1 && !['the', 'and', 'for', 'with', 'from', 'this'].includes(t));
+}
+
+/**
+ * Build a simple bag-of-words vector for a document.
+ */
+function bowVector(tokens, vocabulary) {
+  const vec = new Array(vocabulary.size).fill(0);
+  for (const token of tokens) {
+    const idx = vocabulary.get(token);
+    if (idx !== undefined) vec[idx]++;
+  }
+  return vec;
+}
+
+/**
+ * Cosine similarity between two vectors.
+ */
+function cosineSimilarity(a, b) {
+  let dot = 0, normA = 0, normB = 0;
+  for (let i = 0; i < a.length; i++) {
+    dot += a[i] * b[i];
+    normA += a[i] * a[i];
+    normB += b[i] * b[i];
+  }
+  if (normA === 0 || normB === 0) return 0;
+  return dot / (Math.sqrt(normA) * Math.sqrt(normB));
+}
+
+/**
+ * Build a TF-IDF index from extracted symbols.
+ */
+function buildTfIdfIndex(symbols) {
+  // Build vocabulary
+  const vocabSet = new Set();
+  const docs = symbols.map(s => ({
+    tokens: tokenize(`${s.name} ${s.signature} ${s.type} ${s.lang} ${s.file}`),
+    symbol: s,
+  }));
+
+  for (const doc of docs) {
+    for (const token of doc.tokens) vocabSet.add(token);
+  }
+
+  const vocabulary = new Map();
+  [...vocabSet].forEach((word, i) => vocabulary.set(word, i));
+
+  // Compute IDF
+  const df = new Array(vocabulary.size).fill(0);
+  for (const doc of docs) {
+    const seen = new Set();
+    for (const token of doc.tokens) {
+      const idx = vocabulary.get(token);
+      if (idx !== undefined && !seen.has(idx)) {
+        df[idx]++;
+        seen.add(idx);
+      }
+    }
+  }
+
+  const N = docs.length;
+  const idf = df.map(d => d === 0 ? 0 : Math.log((N + 1) / (d + 1)) + 1);
+
+  // Build TF-IDF vectors for each document
+  const vectors = docs.map(doc => {
+    const tf = new Array(vocabulary.size).fill(0);
+    for (const token of doc.tokens) {
+      const idx = vocabulary.get(token);
+      if (idx !== undefined) tf[idx]++;
+    }
+    // TF normalization
+    const maxTf = Math.max(...tf);
+    if (maxTf > 0) {
+      for (let i = 0; i < tf.length; i++) {
+        tf[i] = 0.5 + 0.5 * (tf[i] / maxTf);
+      }
+    }
+    return tf.map((v, i) => v * idf[i]);
+  });
+
+  return { vocabulary, idf, vectors, docs };
+}
+
+/**
+ * Search the index for symbols relevant to the query.
+ */
+function searchIndex(index, query, topK = 10) {
+  const queryTokens = tokenize(query);
+  if (queryTokens.length === 0) return [];
+
+  const queryVec = bowVector(queryTokens, index.vocabulary);
+
+  // IDF weight the query
+  for (let i = 0; i < queryVec.length; i++) {
+    queryVec[i] *= (index.idf[i] || 1);
+  }
+
+  // Score all docs
+  const scores = index.vectors.map((vec, i) => ({
+    score: cosineSimilarity(queryVec, vec),
+    symbol: index.docs[i].symbol,
+  }));
+
+  // Sort and return top K
+  return scores
+    .filter(s => s.score > 0)
+    .sort((a, b) => b.score - a.score)
+    .slice(0, topK);
+}
+
+/**
+ * Build the codebase index.
+ *
+ * @param {string} repoRoot - Root directory of the repository
+ * @param {object} [options]
+ * @param {string[]} [options.includePatterns] - Glob patterns for files to include
+ * @param {string[]} [options.excludePatterns] - Glob patterns for files to exclude
+ * @param {number} [options.maxFiles=500] - Maximum number of files to index
+ * @returns {object} Index object with symbols, stats, and search capability
+ */
+function buildIndex(repoRoot, options = {}) {
+  const maxFiles = options.maxFiles || 500;
+  const startTime = Date.now();
+  const allSymbols = [];
+  const filesScanned = [];
+  let filesProcessed = 0;
+
+  // Ensure index directory exists
+  const indexDir = path.join(repoRoot, INDEX_DIR);
+  if (!fs.existsSync(indexDir)) {
+    fs.mkdirSync(indexDir, { recursive: true });
+  }
+
+  // Walk the directory tree
+  function walk(dir, relativePath) {
+    if (filesProcessed >= maxFiles) return;
+
+    let entries;
+    try {
+      entries = fs.readdirSync(dir, { withFileTypes: true });
+    } catch {
+      return; // Skip unreadable directories
+    }
+
+    for (const entry of entries) {
+      if (filesProcessed >= maxFiles) return;
+
+      const fullPath = path.join(dir, entry.name);
+      const relPath = relativePath ? `${relativePath}/${entry.name}` : entry.name;
+
+      if (entry.isDirectory()) {
+        if (!SKIP_DIRS.has(entry.name) && !entry.name.startsWith('.')) {
+          walk(fullPath, relPath);
+        }
+      } else if (entry.isFile()) {
+        const ext = path.extname(entry.name).toLowerCase();
+        if (INDEXABLE_EXTS.has(ext)) {
+          try {
+            const source = fs.readFileSync(fullPath, 'utf-8');
+            // Skip very large files (> 500KB)
+            if (source.length > 500 * 1024) return;
+
+            const symbols = extractSymbols(source, relPath);
+            allSymbols.push(...symbols);
+            filesScanned.push(relPath);
+            filesProcessed++;
+          } catch {
+            // Skip unreadable files
+          }
+        }
+      }
+    }
+  }
+
+  walk(repoRoot, '');
+
+  const tfidfIndex = buildTfIdfIndex(allSymbols);
+
+  const stats = {
+    filesScanned: filesProcessed,
+    symbolsExtracted: allSymbols.length,
+    timeMs: Date.now() - startTime,
+    languageBreakdown: {},
+  };
+
+  for (const s of allSymbols) {
+    stats.languageBreakdown[s.lang] = (stats.languageBreakdown[s.lang] || 0) + 1;
+  }
+
+  const index = {
+    version: 1,
+    createdAt: new Date().toISOString(),
+    repoRoot,
+    stats,
+    symbols: allSymbols,
+    tfidf: {
+      // Store just what we need for search
+      vocabulary: [...tfidfIndex.vocabulary.keys()],
+      idf: tfidfIndex.idf,
+      vectors: tfidfIndex.vectors,
+      docs: tfidfIndex.docs.map(d => ({ symbol: d.symbol })),
+    },
+  };
+
+  // Persist to disk
+  try {
+    fs.writeFileSync(path.join(indexDir, INDEX_FILE), JSON.stringify(index, null, 2), 'utf-8');
+    fs.writeFileSync(path.join(indexDir, META_FILE), JSON.stringify({
+      lastBuilt: new Date().toISOString(),
+      filesScanned: filesProcessed,
+      symbolsExtracted: allSymbols.length,
+    }, null, 2), 'utf-8');
+  } catch {
+    // Non-fatal: index persists in memory even if write fails
+  }
+
+  return index;
+}
+
+/**
+ * Load an existing index from disk.
+ *
+ * @param {string} repoRoot - Root directory of the repository
+ * @returns {object|null} Index object or null if not found
+ */
+function loadIndex(repoRoot) {
+  const indexPath = path.join(repoRoot, INDEX_DIR, INDEX_FILE);
+  try {
+    if (!fs.existsSync(indexPath)) return null;
+    const data = JSON.parse(fs.readFileSync(indexPath, 'utf-8'));
+    // Rebuild vocabulary Map from stored array
+    if (Array.isArray(data.tfidf) && Array.isArray(data.tfidf.vocabulary)) {
+      data.tfidf.vocabulary = new Map(data.tfidf.vocabulary.map((w, i) => [w, i]));
+    } else if (data.tfidf && Array.isArray(data.tfidf.vocabulary)) {
+      data.tfidf.vocabulary = new Map(data.tfidf.vocabulary.map((w, i) => [w, i]));
+    }
+    return data;
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * Retrieve relevant context for a given diff.
+ *
+ * Given a git diff (which tells us which files changed and what lines),
+ * this function finds related symbols from the codebase index:
+ * 1. Same-file context: symbols defined in the changed files
+ * 2. Cross-file context: imports/exports related to changed files
+ * 3. Semantic context: top-K similar symbols across the codebase
+ *
+ * @param {object} index - The codebase index (from buildIndex or loadIndex)
+ * @param {string} diff - Git diff text
+ * @param {object} [options]
+ * @param {number} [options.maxContext=15] - Max symbols to return
+ * @param {boolean} [options.includeSemantic=true] - Include semantic search results
+ * @returns {object} Context object with symbols grouped by type
+ */
+function retrieveContext(index, diff, options = {}) {
+  if (!index || !index.symbols || index.symbols.length === 0) {
+    return { symbols: [], summary: 'No codebase index available. Run `coderev index` first.' };
+  }
+
+  const maxContext = options.maxContext || 15;
+  const includeSemantic = options.includeSemantic !== false;
+
+  // Extract changed file paths from diff
+  const changedFiles = new Set();
+  const diffLines = diff.split('\n');
+  for (const line of diffLines) {
+    const match = line.match(/^\+\+\+ b\/(.+)/);
+    if (match) changedFiles.add(match[1]);
+  }
+
+  const results = new Map(); // file+name -> symbol dedup
+
+  // 1. Same-file: symbols defined in changed files
+  for (const sym of index.symbols) {
+    if (changedFiles.has(sym.file)) {
+      const key = `${sym.file}:${sym.name}`;
+      if (!results.has(key)) results.set(key, { ...sym, relevance: 'same_file' });
+    }
+  }
+
+  // 2. Cross-file: imports/exports related to changed files
+  // Find modules imported by changed files
+  for (const sym of index.symbols) {
+    if (changedFiles.has(sym.file) && sym.type === 'import' && sym.name) {
+      // Find symbols exported by the imported module
+      for (const other of index.symbols) {
+        if (other.type === 'export' && other.file.includes(sym.name.replace(/^\.?\/?/, ''))) {
+          const key = `${other.file}:${other.name}`;
+          if (!results.has(key)) results.set(key, { ...other, relevance: 'cross_file' });
+        }
+      }
+    }
+  }
+
+  // 3. Semantic search: find similar symbols across codebase
+  if (includeSemantic && results.size < maxContext) {
+    // Build query from diff context (focus on function/class names in diff)
+    const queryParts = [];
+    for (const sym of index.symbols) {
+      if (changedFiles.has(sym.file)) {
+        queryParts.push(sym.name);
+        queryParts.push(sym.type);
+      }
+    }
+    const query = queryParts.join(' ') || diffLines.slice(0, 20).join(' ');
+
+    // searchIndex expects the flat TF-IDF structure with Map vocabulary
+    const tfidf = index.tfidf || index;
+    // Ensure vocabulary is a Map (may be stored as array in JSON)
+    if (Array.isArray(tfidf.vocabulary)) {
+      tfidf.vocabulary = new Map(tfidf.vocabulary.map((w, i) => [w, i]));
+    }
+    const semanticResults = searchIndex(tfidf, query, maxContext);
+    for (const r of semanticResults) {
+      if (changedFiles.has(r.symbol.file)) continue; // Already have same-file
+      const key = `${r.symbol.file}:${r.symbol.name}`;
+      if (!results.has(key)) {
+        results.set(key, { ...r.symbol, relevance: 'semantic', score: r.score.toFixed(3) });
+      }
+    }
+  }
+
+  // Convert to array, limit
+  const contextSymbols = [...results.values()].slice(0, maxContext);
+
+  // Group by relevance
+  const grouped = {
+    sameFile: contextSymbols.filter(s => s.relevance === 'same_file'),
+    crossFile: contextSymbols.filter(s => s.relevance === 'cross_file'),
+    semantic: contextSymbols.filter(s => s.relevance === 'semantic'),
+  };
+
+  return {
+    symbols: contextSymbols,
+    grouped,
+    changedFiles: [...changedFiles],
+    totalIndexed: index.symbols.length,
+    summary: formatContextSummary(grouped, changedFiles),
+  };
+}
+
+/**
+ * Format context as a compact text block for prompt injection.
+ */
+function formatContextSummary(grouped, changedFiles) {
+  const parts = [];
+
+  if (changedFiles.length > 0) {
+    parts.push(`Changed files: ${changedFiles.join(', ')}`);
+  }
+
+  if (grouped.sameFile.length > 0) {
+    parts.push(`\n📄 Same-file context (${grouped.sameFile.length} symbols):`);
+    for (const s of grouped.sameFile.slice(0, 10)) {
+      parts.push(`  ${s.type}: ${s.name}${s.signature ? `(${s.signature})` : ''} (${s.file}:${s.line})`);
+    }
+  }
+
+  if (grouped.crossFile.length > 0) {
+    parts.push(`\n🔗 Related symbols (${grouped.crossFile.length}):`);
+    for (const s of grouped.crossFile.slice(0, 5)) {
+      parts.push(`  ${s.type}: ${s.name} in ${s.file}`);
+    }
+  }
+
+  if (grouped.semantic.length > 0) {
+    parts.push(`\n🔍 Similar symbols across codebase (${grouped.semantic.length}):`);
+    for (const s of grouped.semantic.slice(0, 5)) {
+      parts.push(`  ${s.type}: ${s.name} in ${s.file}`);
+    }
+  }
+
+  return parts.join('\n');
+}
+
+/**
+ * Build a context string for injection into review prompts.
+ * This is the main integration point with reviewer.js
+ *
+ * @param {object} index - The codebase index
+ * @param {string} diff - Git diff text
+ * @param {object} [options]
+ * @returns {string} Context string for prompt injection
+ */
+function buildReviewContext(index, diff, options = {}) {
+  if (!index) return '';
+
+  const ctx = retrieveContext(index, diff, options);
+  if (!ctx.symbols || ctx.symbols.length === 0) return '';
+
+  let contextBlock = `
+## 📚 Codebase Context (Retrieved via RAG)
+
+The following symbols were found in the codebase that may be relevant to this change:
+
+`;
+
+  // Same-file symbols (most important)
+  if (ctx.grouped.sameFile.length > 0) {
+    contextBlock += `### Same File Symbols\n`;
+    for (const s of ctx.grouped.sameFile.slice(0, 8)) {
+      contextBlock += `- \`${s.type}\` **${s.name}**${s.signature ? `(${s.signature})` : ''} at line ${s.line}\n`;
+      if (s.snippet && s.snippet.length < 300) {
+        contextBlock += `  \`\`\`\n${s.snippet}\n  \`\`\`\n`;
+      }
+    }
+  }
+
+  // Cross-file references
+  if (ctx.grouped.crossFile.length > 0) {
+    contextBlock += `### Cross-File References\n`;
+    for (const s of ctx.grouped.crossFile.slice(0, 5)) {
+      contextBlock += `- \`${s.type}\` **${s.name}** in \`${s.file}\`\n`;
+    }
+  }
+
+  // Semantic matches
+  if (ctx.grouped.semantic.length > 0) {
+    contextBlock += `### Semantically Similar\n`;
+    for (const s of ctx.grouped.semantic.slice(0, 5)) {
+      contextBlock += `- \`${s.type}\` **${s.name}** in \`${s.file}\`\n`;
+    }
+  }
+
+  contextBlock += `\nUse this context to understand call chains, type relationships, and coding patterns.`;
+
+  return contextBlock;
+}
+
+/**
+ * Check if an index needs rebuilding (stale or non-existent).
+ *
+ * @param {string} repoRoot - Repository root
+ * @param {number} [maxAgeHours=24] - Max age in hours before considered stale
+ * @returns {boolean}
+ */
+function isIndexStale(repoRoot, maxAgeHours = 24) {
+  const metaPath = path.join(repoRoot, INDEX_DIR, META_FILE);
+  try {
+    if (!fs.existsSync(metaPath)) return true;
+    const meta = JSON.parse(fs.readFileSync(metaPath, 'utf-8'));
+    const builtTime = new Date(meta.lastBuilt).getTime();
+    return (Date.now() - builtTime) > maxAgeHours * 60 * 60 * 1000;
+  } catch {
+    return true;
+  }
+}
+
+module.exports = {
+  buildIndex,
+  loadIndex,
+  retrieveContext,
+  buildReviewContext,
+  extractSymbols,
+  searchIndex,
+  buildTfIdfIndex,
+  isIndexStale,
+  INDEX_DIR,
+  INDEXABLE_EXTS,
+  SKIP_DIRS,
+};